Is there a correlation between payload length and potential packet loss?

End Devices (Nodes)
1

Hi,

thinking about packet loss in LoRa, I came up with a question for which, after lots of reading, I didnt find an answer:

Are packets with longer payloads more likely to get lost? E.g. is there a X% higher chance that a packet with a 40 byte long payload is lost compared to a packet with 10 byte payload? Of course assuming that all other parameters (RSSI, SNR, SF) are the same.

Thanks & best regards
Philipp

2

Hi Phillipp

Yes it is true that the longer the airtime of a packet the bigger the chance that a packet suffers from interference. This is one of the measures in fighting interference in the ISM bands. Minimimize the airtime by reducing payload size end increasing datarate.

3

Hi @PhilippRe this is especially true for moving/mobile nodes as there is increased risk of LOS obstruction causing loss of end of what would otherwise be a successful packet Tx. e.g. many GPS Trackers (which likely also Tx other sensor data e.g battery or temp etc.) are on air for 1.3-1.7s @ SF12 and at speeds >100kmph have have traversed ~50-75m tangentially to any receiving GW giving plenty of time/distance to fall into an obstructions shadow. (Note also that for mobile devices esp if moving perpendicular to GW speed of movement - especially higher speeds at higher SF’s - becomes a potential issue due to SIgnal Cohesivity concerns, causing potential packet loss. IIRC whilst SF7 can be resilient to speeds >180kpm, at SF12 speeds as low as 45kph or even ~25kph ‘can’ be a problem!). This can be frustrating in e.g. tracking applications where you typically want to follow a device at significant distance and between GW’s…implying use of higher SF’s - 10, 11 & 12! :frowning:

2 Likes
4

While the two other remarks are correct, there is also the simple fact that at a given SNR the bit error rate is the same but the packet error rate will increase since there is more bits. The formula is PER = 1 - (1-BER)^n where n is the number of bits in the packet.

2 Likes
5

While a long packet is more likely to have an error, at the same time, you don’t want an overly short packet as in LoRaWAN you are paying something like 13 bytes overhead just for the header and trailing MIC before you’ve moved any useful data.

2 Likes
6

This I’ve been looking for! Thank you! Do you have a source/reference for this formula?

7

It is “just” probability:

  • the BER is the probability of having one bit in error.
  • If you have a packet with only one bit, the probability of having no error is 1-BER.
  • If you have n bit in a packet the probability to have no bit in error (1-BER) ^ n
  • This gives the probability of getting a packet received correctly, so to get the PER you just need to do 1-(1-BER)^n
1 Like
8

Ah yes, of course. That makes sense. Thank you very much!